Releases | Open Source Science

What's Changed

changes to future seed by @graemeblair in https://github.com/DeclareDesign/DeclareDesign/pull/494

Full Changelog: https://github.com/DeclareDesign/DeclareDesign/compare/1.0.6...1.0.8

- R
Published by graemeblair almost 2 years ago

- R
Published by graemeblair almost 2 years ago

- R
Published by graemeblair over 2 years ago

New tidy(diagnosis) function to construct a tidy data frame of diagnosand estimates and summary statistics including bootstrapped standard error and confidence intervals.
Remove drawassignment and drawsample function.
Added diagnosis duration to summary function.
Bug fixes.

- R
Published by graemeblair almost 4 years ago

To simplify output of diagnoses, we changed the names of the variables from designlabel to design, inquirylabel to inquiry, and estimator_label to estimator.
declareassignment() and declaresampling() have the default values for legacy set to FALSE. You can still use the legacy versions of these functions by manually setting legacy = TRUE for some time, but this functionality will later be removed.

- R
Published by graemeblair over 4 years ago

Rapid development phase is beginning to prepare for DeclareDesign 1.0.
Add new step declare_model for defining the model of the world including sample size, levels of the data, and variables.
Add new step declare_inquiry to replace declare_estimand. declare_estimand is still available but deprecated.
Soft-introduce new syntax for declareassignment and declaresampling. Old syntax is still available with legacy = TRUE, the current default. To use the new syntax, set legacy = FALSE. In future versions of DeclareDesign, the default will be set to FALSE.
reveal_outcomes, created in 0.24.0, has been removed as a step.
Change labels produced in diagnoses and run_design output to be inquiry_label rather than estimand_label.
Allow estimands to be functions of other estimands.

- R
Published by acoppock almost 5 years ago

Add new step declare_measurement for measuring outcome variables.
Add declare_test to enable hypothesis testing where no estimand is targeted. For example, declare_test could be used for a K-S test of distributional equality and declare_estimator for a difference-in-means estimate of an average treatment effect.
Add model_summary option to declare_estimator, to enable specifying a model and then a separate post-estimation function to extract coefficient estimates (e.g., estimate of a treatment effect) or model summary statistics (e.g., R^2 or the result of an F-test from a regression).
Simplify declare_diagnosands functionality. diagnose_design() by default runs an internal function with a set of default diagnosands, including power, RMSE, bias, type S rate, coverage, mean estimate, and mean estimand.
Improve compatibility with dplyr verbs as handlers. filter now works.
Rename declare_reveal to reveal_outcomes. Both continue to work.

- R
Published by graemeblair over 5 years ago

Fix ability to set samplingvariable in declaresampling.
Add ability to retain nonsampled data after sampling via dropnonsampled flag in declaresampling.

- R
Published by graemeblair almost 6 years ago